The context
People do not shop for clothes in keywords. They look for “a light jacket for a summer wedding” or “something like this, but in navy”. Traditional e-commerce search matches the words in a query to the words in product titles and tags, and when the two do not line up, relevant products never appear.
For a fashion retailer, every search that returns the wrong results is a visitor who leaves without buying.
The challenge
The business needed search that understood what shoppers meant: the style, the occasion and the feel of what they were looking for, and which products in a large, constantly changing catalogue matched it. It had to be fast enough for a live store and maintainable as the catalogue changed every season.
What we did
We built an AI search engine that represents both shoppers’ requests and products by meaning rather than by keywords. Each product is described by its style, attributes and appearance; each query is interpreted for intent. Search then returns the products closest in meaning, so “a light jacket for a summer wedding” finds linen blazers even if no product is tagged with those words.
We engineered it for production: quick to respond, updated as new products arrived, and integrated into the client’s existing store.
The result
The search engine went live in production, and the client acquired the intellectual property, a clear signal of its value to the business. A paid engagement to support and develop it continued afterwards.
Why it worked
- It matched how customers actually think. Searching by meaning closed the gap between what shoppers type and how products are described.
- It was built to run, not to demo. Speed, catalogue updates and integration were treated as requirements from day one.